Claude now has a real identity-verification gate: Anthropic’s own help page says some users can be asked for a government-issued photo ID, device camera access, and, in some cases, a live selfie, with checks run by Persona. The important part is scope: Engadget reports Anthropic says this applies only in a “small number of cases”, not as a blanket ID rule for everyone using Claude.
That makes this better understood as selective access control, not universal KYC. Anthropic says identity checks can be used for “certain capabilities,” “routine integrity checks,” and “other safety and compliance measures”, which places the feature inside trust and enforcement workflows rather than normal sign-up friction.
Identity-verification prompts are documented, selective, and tied to access control
Anthropic’s identity-verification help page is unusually specific about the mechanics. If Claude asks for verification, the process can require three things: a valid government-issued photo ID, access to your device camera, and possibly a live selfie. Anthropic says Persona handles that flow.
What the page does not say is almost as important. It does not quantify how many users are affected, which Claude plans or surfaces trigger it, or whether any particular advanced feature always requires ID. Anthropic says the checks are for “a few use cases”; that is evidence of selectivity, not breadth.
The strongest independent narrowing comes from Engadget’s report, which says an Anthropic spokesperson described the prompts as applying in a “small number of cases” linked to activity indicating potentially fraudulent or abusive behavior. That is still Anthropic’s characterization, not an external audit, but it is the clearest available evidence against the idea that Claude now demands ID from everyone.
Anthropic also ties the verification step directly to enforcement outcomes. Its help page says users may be asked to verify identity after repeated usage-policy violations, terms-of-service violations, account creation from an unsupported location, or use by someone under 18. In other words, this is not just “prove you are real.” It is also “prove you are eligible to keep access.”
That matters because Anthropic already has documented account actions around restrictions and bans, which makes the ID step fit naturally into broader Anthropic bans and enforcement workflows. The shift is subtle but real: access to a frontier model is starting to look less like ordinary software login and more like a gated service with identity-linked checks.
Anthropic says the verification data is used only to verify identity and “not for training AI models”. TechRadar’s report, which tracks closely with the help-page language, also notes Anthropic’s statement that verification is for selected capabilities and checks rather than general model training.
Age assurance uses a separate Yoti flow with pass/fail results
Anthropic’s age-assurance page describes a different system. If Anthropic suspects a user may be under 18, it says the user may be asked to complete age assurance through Yoti, using one of several methods: facial age estimation, ID verification, or the Yoti Digital ID app.
The key distinction is what Anthropic says it receives back. For age assurance, Anthropic says it gets only a pass/fail result from Yoti, not the underlying age-estimation details. That is much narrower than a general identity-verification flow.
A quick side-by-side makes the split clearer:
| Flow | What Anthropic says it is for |
|---|---|
| Identity verification via Persona | Certain capabilities, integrity checks, and safety/compliance measures |
| Age assurance via Yoti | Checking whether a user is 18 or older when Anthropic suspects they may be underage |
The two systems can look similar from the user side because both may involve a selfie or an ID document. But they are not the same policy. Anthropic’s own help pages separate them, and collapsing them into one “Claude now scans everyone’s face” story would be wrong.
Verification data handling runs through Persona and Yoti, not model training
Anthropic’s identity-verification page says Persona processes the verification, while the age-assurance page says Yoti handles age checks. Anthropic’s public positioning is that these are vendor-run verification layers attached to account access, not training inputs for Claude itself.
Biometric Update adds useful but secondary detail, reporting that Anthropic’s privacy language can include items such as ID images, date of birth or ID number, photos or videos, facial geometry templates, and verification results, depending on the verification method. That should be read as possible data categories around verification, not proof that every Claude user is handing over all of them.
This is also where Claude’s broader privacy posture matters. Anthropic has separately described Claude privacy controls, but identity verification sits outside the usual “will my prompts train the model?” conversation. It is closer to eligibility infrastructure: the data exists to decide who can use what, and under what conditions.
Forrester frames that as part of a wider move toward identity-backed access controls for AI systems. That broader thesis is plausible, but here the strongest evidence remains narrower and simpler: Anthropic has documented a selective ID gate, named the vendors, listed the possible checks, and tied the process to capability access and enforcement.
Claude now has a real identity-verification gate: Anthropic’s own help page says some users can be asked for a government-issued photo ID, device camera access, and in some cases a live selfie, with checks run by Persona.
The next thing to watch is whether Anthropic expands the help-page language into product-level detail, for example, which Claude features, plans, or risk states actually trigger the prompts. As of now, the company has documented the mechanism, but not the rollout size.
Key Takeaways
- Claude can require identity verification using a government ID, camera access, and sometimes a live selfie, with Persona running the checks.
- The scope appears selective, not universal, because Anthropic says the feature is for “a few use cases” and Engadget reports it applies in a “small number of cases”.
- Anthropic ties verification to access and enforcement, including certain capabilities, integrity reviews, safety/compliance measures, unsupported locations, policy violations, and under-18 use.
- Age assurance is a separate flow handled by Yoti, and Anthropic says it receives only a pass/fail result.
- Anthropic says verification data is not used to train Claude’s models and is used solely for verification.
Further Reading
- Identity verification on Claude, Anthropic’s primary help page on Persona-run ID checks, accepted documents, and enforcement triggers.
- Age assurance on Claude, Anthropic’s separate help page on Yoti-based age checks and pass/fail handling.
- Anthropic will ask Claude users to verify their identities ‘for a few use cases’, Independent report on Anthropic’s claim that the prompts affect only a small number of cases.
- Update on identity, age verification for Claude prompts user pushback, Trade coverage of verification-data categories and policy language.
- Anthropic’s Claude Rolls Out End-User Identity Verification, Analyst context on what selective AI identity gating may mean.
Last reviewed: 2026-06